R3 Observations - Image quality

Ephemeris wrote:

Ephemeris wrote:

RLight wrote:

Ephemeris wrote:

RLight wrote:

Ephemeris wrote:

RLight wrote:

Ephemeris wrote:

RLight wrote:

Ephemeris wrote:

Thanks for the extra information.

Id like to see some controlled images to show fine detail as the ISO setting increases.

The DPR studio does this, somewhat.

I made this point as a route to not using that resource specifically. Our own tests of the R6 Vs R5 meant that the r6 was effectively not suitable at all, and a backward step on the R we had before. Lots of talk of low light performance but we are interested in signal, detail at high ISO. The R5 was a clear winner of the 3 bodies. Have never tried an R3.

We use R5s for this purpose so if the thought that the R3 can provide more detail than the R5 I'd love to see that and maybe we have an alternative.

Using the DPReview comparison tool, R Vs 5Ds Vs R3 Vs R5 at 12800 (limited by 5Ds) the R3 does t look to show the detail of the others. However any other information greatly received.

45MP is more than 24MP; sheer detail rendition the R5 "wins"

There is more to the story, though. Let me draw your attention to the fine print of the DPR studio comparison itself first...

Note, no brightness correction, and the RF 50mm f/1.2L is employed

Note, this one is brightness corrected, and uses the RF 85mm instead to standardize on the newer 85mm standard.

The RF 85mm and 50mmL's have differing light transmission, vignetting and resolution differences at the center and fringes of the image making the comparison different from an input perspective.

Likewise, note the brightness correction application disclaimer on the R3 RAW. Why? The answer is in the ISO standard itself...

https://www.dxomark.com/Cameras/Compare/Side-by-side/Canon-EOS-R3-versus-Canon-EOS-R5___1367_1355

At a Mfrg ISO of 400 (ISO set in-camera), the R3 delivers an ISO sensitivity of 285 vs the R5's ISO sensitivity of 248; different light sensitivity and thereby necessitating brightness correction to have perceptual equity necessary for the DPR studio's "one size fits all" approach. However in practice, you might need say ISO500 on your R5 to have the same exposure as ISO400 on the R3 making the DPR studio approach somewhat misleading when brightness correction has to be applied for a true apples to apples comparison of the effects of noise handling of the two.

Comparing the R3 to the R6 (original) on the other hand, you see the ISO sensitivities are neck and neck

https://www.dxomark.com/Cameras/Compare/Side-by-side/Canon-EOS-R6-versus-Canon-EOS-R3___1354_1367

Now this isn't the end of the story, I've just demonstrated that bigger pixels absorb more light. There's also BSI going on in the R3 via the stacked sensor... Courtesy of TDP, which uses Canon's own DPP4 for his studio samples, not Adobe, and with no corrections, a 24MP FSI (R6 II) vs the 24MP BSI (R3) can be accurately compared to demonstrate the SnR effects of an FSI sensor vs BSI via the link below to see the reduced noise as a result of BSI... Feel free to play around with different ISO settings. Watch the whites and blacks in particular and note the reduced Chroma and Luminance noise on the R3 vs the R6 II, without having to account for resampling (identical resolution)

https://www.the-digital-picture.com/Reviews/Camera-Noise.aspx?Camera=1553&Test=0&ISO=6400&CameraComp=1633&TestComp=0&ISOComp=6400

This explains why the R6 "beats" the R5 in terms of low light sensitivity for AF, and higher ISO range, which according to Canon's literature, gives an indication of better low light potential. And the latter explains why the R3 bests the R6 (and R6 II), as it should being a stacked (BSI) sensor.

Our work has told us this R6 wasn't better, and actually didn't provide as much usable signal as this R.

We only looked at the R6 as a second body becuase the R5 was almost impossible to get hold of. We decided to wait and live with one body.

You can use the above TDP link to compare the R5 against the R3, however, TDP's comparison does not provide for resampling to give say an apples to apples of a various higher ISOs without brightness compensation e.g. DPR Studio. DPR studio permits resampling on the other hand, but can misrepresent ISO performance.

What I can say, and will restate, the R5 will deliver more resolution at base ISO (100) than the R3 at ISO100. 45MP is 45MP and 24MP is 24MP after all. Where things get murky in regards to only detail (thus exempting color and dynamic range from consideration) rendition is higher ISOs, say 12800. I can say the R5 in my experience, delivers more (not better, more)

More is better that's the point.

detail, even at higher resolutions. But, at the cost of color, tones and dynamic range. The R3 "wins" in those just as the R5 "wins" with regard to resolution only, as it should (45MP is 45MP).

This is why I stated, studio and landscape shooters should look at the R5. Event and wedding shooters should look at the R3 (or R6 Mark II). If both? You either have to choose, or bring the best tool for the occasion.

Could you show a test (which is what I asked) that shows some images, with detail, similar to the studio shot at ISO12800 and above (like I used the studio tool to include the 5Ds which looked to well outperform the R3) that meets your test criteria? I'm looked for usable signal detail comparisons and where one can resolve and one cannot.

Now for say sports shooters, or in the middle (mixed shooting)? I'm going to nod Canon's recommendation; if you need to crop, R5. If you don't? R3. To quote the first bullet point of the R5...

"Can shoot higher resolution still images and videos due to its higher megapixel count, which provides more flexibility if you need to crop... As such, it is the better choice if post-processing and retouching is a significant part of your workflow."

Cropping, is a part of post-processing workflow...

With regard to the R3, I'll quote them again...

"This makes it a camera geared towards capturing moments. If your shooting style revolves around capturing decisive moments that occur right before your eyes, the fast EOS R3 will be a good choice."

Sounds like it doesn't have suitable resolution?

https://snapshot.canon-asia.com/article/eng/eos-r3-vs-eos-r5-which-one-should-i-choose

This all comes down to choosing the right tool for the job. I'll say it bluntly, again; wedding shooters, photojournalists, event shooters? R3 (or R6 Mark II for a fraction of the price, albeit you loose the specialities of the R3). Landscape shooters, studio shooters? R5.

Events, engineering shots indoor and outdoor, macro, 8k video. Why would I use an R3? We can't live with that resolution and until I can see that it can provide the detail that the 5Ds, R, R5 can then why would it be helpful to me?

That's my point back at the start. I'm not having a moan at the R3 in the slightest, I bet if love to shoot with one if I could afford it as a hobby. It has some fabulous features I don't have. But maybe the R5 replacement will also?

R1; but, this means a 45MP+ sensor, with Stacked or Global.

I don't follow why the R1 is mentioned as it doesn't exist and a hypothetical discussion is possibly not that valuable.

Because Canon will do a 45MP sensor (or more) which can resolve 8K, and it'll be stacked or global. All to say you're going to have the same challenges even when a R5 successor emerges.

Well it's already 8k and 45MP. Rolling shutter is decent compared to its rivals this conversation is about the R3 and that in struggling to see when it's going to resolve more information. If it can then I would give it a try.

The R5 successor, whatever that may be is a wait and see.

I'm still not seeing the advantage for our work of the R3. I don't see it producing more detail, even than the R mostly.

I agree. The R3 and R6 II are on par with the R in resolving power. You're not missing the point. You'd need to punch say 16000 on an R5 for say an R3 to pull ahead, in my opinion, in sheer detail, not color or tones. The R3 pulls ahead of the R5 in terms of color easily early on though.

They look behind from what I can see. The R6 was behind the R by a good step for us hence we waited for another r5. Can the R3 jump up here? That's my question.

As for colour, again im not sure becuase I don't have one to compare. I'd be surprised however but I and I'm sure others be interested. Maybe the wildlife folk could benefit and swap from an R5?

At 12800 the studio shots, which is all I have to compare look behind the 5Ds. I'm not saying it's a perfect comparison hence asking for another comparison.

That means the former stays the same (ISO sensitivity; smaller pixels) but the SnR improves due to reduced noise from BSI. All to say the R1 will be "better" from the R5, but there's still going to be times where an R3 is a better tool than a R1 potentially.

It's a pretty darn good camera the R5 and the weak part of the chain in my shop is the user. Still learning how to get more from it.

I'm going to do an out-thoughts section of this mini series that I plan to touch on this. Folks like yourself are asking these questions, and I've asked them myself.

Maybe some 102400 ISO setting and underexposed by 3 stops the R3 can pull more detail? But can it? Really?

No, it can pull more detail in my book by 16000. But, it can pull more colors and tones by 800, with less detail.

I'm not sure I really understand what that means. The amplitude of the colour? Colour accuracy possibly? (I see your looking at DxO bits per channel from a RAW file)

Yes, color accuracy. As your SnR diminishes, it's harder to interpret the signal.The aggregation of the pixels permits the R5 to hold unto a resolution advantage, despite having poor-er ISO performance, though, but color takes a hit. This shows up in skin tones, and "pop" an image has. You really have to look at a ton of samples or shoot yourself to get the feel for this. There's being able to read someone's newspaper text, and then there's the ambiance an event has. The R5 wins with reading a newspaper, the R3 wins in terms of gold staying gold and not turning white or grey, or pink staying pink, and not turning white. Etc. The quality of light, improves. Things like the color of someone's shirt under artificial light, at say ISO12800, may, and will differ between an R5 and R6 (and R3, R6 Mark II); dark blue may presented as blue and the like.

Dustin Abbott's review touches on this... Recommend a read...

https://dustinabbott.net/2021/12/canon-eos-r3-review/

https://dustinabbott.net/wp-content/gallery/canon-eos-r3-review/32-R5-ISO-Comparison.jpg

Note the saturation on the reds.

I did find a good reference for true detail vs ISO performance...

https://www.photoreview.com.au/reviews/mirrorless-cameras/mirrorless-cameras-full-frame/canon-eos-r3/

https://www.photoreview.com.au/reviews/mirrorless-cameras/mirrorless-cameras-full-frame/canon-eos-r5/

The R5 maintains resolution advantage, period, throughout the spectrum on paper. The author of that one used an RF 70-200 f/2.8L, which is arguably a better "bar" than the primes DPR uses which will outright favor higher resolution sensors. It's a more fair test. A good example as to why, my favorite lens, the RF 28-70 f/2L, doesn't benefit as much from 45MP as say a RF 50mm f/1.2L...

https://www.opticallimits.com/canon_eos_ff/1139-canonrf2870f2?start=1

This comes down to what medium you're outputting to your clientele and what ISOs you're generally shooting. If you're putting out say large prints constantly? R5 all the way. If say Smartphone, and you're hitting ISO12800 frequently? R3.

Well, usually it's purely digital but enlarging to as much detail as possible.

What about the concept that the R3 is a bit noiser due to its sensor design? Is this correct?

Less noisy. R3 uses BSI, you can see in the TDP comparison tool this in effect vs the same 24MP, not BSI R6 Mark II.

I was taking information from the PDR measurements which may not be the same thing you are describing. I recall the R3 has some change in PDR around ISO1250 but below that the R5 has a benefit.

To put this more simply; the R5 "wins" across the board prior to ISO800. It has identical or better color depth, until 800. Afterwards the R3 pulls ahead meaning you're trading better detail rendering for poorer color rendering on the R5 past 800. I agree not just with the numbers, but in practice. See below.

Iis this from DXO? I can only see part of the graph on my phone. What are the three horizontal dotted lines? (I can see it now on DxOs site and the dotted lines)

Yes, you can go to the website to see the metrics yourself, and would encourage you to. I linked it for that reason.

I hope my comments have helped you to understand I did go and look as I couldn't see the graph fully. However I don't know how it's measured which would be nice to know. I'm assuming it a geometric method but not sure. We normally produce these on charts to show the actual measured value and the target value because ensure measurement has error. This may be different of course.

Does he have a conversion problem with the R5?

Cant speak to that. I’ll say the data on DXO mirrors results from Bills photons analysis. I think Bill uses DXO too though.

Your Australian link is more similar to how we measure but we measure thousands of points but it's the same method. They are measuring converted files another to JPEG or TIFF so one must assume the converter has a part to play here and it's not the actual recorded values from the camera?